Haian Huang(深度眸)
Haian Huang(深度眸)
@lluo-Desktop 我们今天发布了 swin-b 和 swin-l 权重，如果有兴趣可以看看 https://github.com/open-mmlab/mmdetection/pull/11458
@lluo-Desktop 可以基于 dev-3.x 重新跑一下，dev-3.x 里面修复了一个数据增强 bug
@christiano12345 I tried it and there was no issue, can you please try it again?
不会呀，我就是用的 3090 训练的，swin-t 的所有实验都是在 3090 上跑完的。
你可以看一下我们发布的训练log
@zkyseu log 里面显示的 memory 是会偏小的，但是 nvdia-smi 中显示是最大能占用的，其实都不准确。3090 应该不会 OOM 的，我们模型就是 3090 上训练的。
0.5:0.5 is mean 0.5 only
@cmjkqyl This could be because the two scripts have different calculation processes. I suggest you rely on the COCO mAP (mean Average Precision) as the reference metric.
@HePengguang Have you updated your code and configurations to the latest 3.1.0 ?
@flytocc Thank you very much. I would like to confirm why a previous pull request (PR) could not align the precision. Was something incorrect there?